CN115908170B - Noise reduction method and device for binocular image, electronic device and storage medium - Google Patents
Noise reduction method and device for binocular image, electronic device and storage medium Download PDFInfo
- Publication number
- CN115908170B CN115908170B CN202211378346.6A CN202211378346A CN115908170B CN 115908170 B CN115908170 B CN 115908170B CN 202211378346 A CN202211378346 A CN 202211378346A CN 115908170 B CN115908170 B CN 115908170B
- Authority
- CN
- China
- Prior art keywords
- image
- view
- euclidean distance
- guiding
- parallax
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 230000009467 reduction Effects 0.000 title claims abstract description 53
- 238000003860 storage Methods 0.000 title claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims abstract description 161
- 238000001914 filtration Methods 0.000 claims abstract description 70
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 57
- 238000012545 processing Methods 0.000 claims description 28
- 238000004590 computer program Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 6
- 238000005516 engineering process Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000012937 correction Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000011946 reduction process Methods 0.000 description 3
- 230000003321 amplification Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000003199 nucleic acid amplification method Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 210000002784 stomach Anatomy 0.000 description 1
- 238000002834 transmittance Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Landscapes
- Image Processing (AREA)
Abstract
The application relates to a noise reduction method, a device, an electronic device and a storage medium for binocular images, wherein the method comprises the following steps: calculating to obtain a first horizontal parallax matrix corresponding to the shallowest layer guiding image of the binocular image by using a non-local mean filtering algorithm; according to the first horizontal parallax matrix, calculating a first Euclidean distance of the first left-view guiding image in the first right-view guiding image and a second Euclidean distance of the first right-view guiding image in the first left-view guiding image by utilizing a non-local mean filtering algorithm; calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by using a non-local mean filtering algorithm; and obtaining a noise reduction image of the corresponding binocular image according to the first Euclidean distance, the second Euclidean distance, the third Euclidean distance and the fourth Euclidean distance. The application realizes the technical effect of reducing the image noise of the binocular image acquired by the binocular endoscope.
Description
Technical Field
The present application relates to the field of image processing technologies, and in particular, to a method and apparatus for denoising a binocular image, an electronic device, and a storage medium.
Background
The endoscope is a detection instrument integrating the traditional optical, human engineering, precision machinery, modern electronics, mathematics, software and other technologies. In particular, endoscopes include image sensors, optical lenses, light source illumination, mechanical devices, etc., which can be accessed orally into the stomach or through other natural tunnels to enable visualization of lesions that cannot be displayed by X-rays. The image obtained by the endoscope generally needs to be subjected to noise reduction treatment before analysis.
The noise reduction method of the endoscope image in the related art is generally a noise reduction algorithm using a common camera image, such as a spatial domain noise reduction method, a frequency domain noise reduction method, a spatial domain and frequency domain combined dual domain noise reduction method, and a deep learning method.
Currently, in the stereoscopic display technology for an endoscope, a polarized display is generally used, and the overall brightness of an image is reduced due to poor light transmittance of the polarized display, so that in order to achieve normal image brightness, an exposure gain parameter of the image needs to be increased, which introduces greater image noise.
At present, no effective solution is proposed for the problem of large noise of images acquired by an endoscope in the related art.
Disclosure of Invention
The embodiment of the application provides a noise reduction method, device, electronic device and storage medium for binocular images, which are used for at least solving the problem that images acquired by endoscopes in the related art are large in noise.
In a first aspect, an embodiment of the present application provides a method for denoising a binocular image, where the method includes:
acquiring binocular images, wherein the binocular images comprise left-view images and right-view images; performing multi-layer downsampling and noise reduction on the binocular image to obtain a multi-layer guide image; calculating to obtain a first horizontal parallax matrix corresponding to the shallowest layer guiding image in the multi-layer guiding images by using a non-local mean filtering algorithm; the shallowest layer guiding image comprises a first left-view guiding image and a first right-view guiding image; according to the first horizontal parallax matrix, a first Euclidean distance of the first left-view guiding image in the first right-view guiding image and a second Euclidean distance of the first right-view guiding image in the first left-view guiding image are calculated by utilizing a non-local mean filtering algorithm; calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by using a non-local mean filtering algorithm; and determining a plurality of similar blocks in the left-view image and the right-view image and weights corresponding to each similar block according to the first Euclidean distance, the second Euclidean distance, the third Euclidean distance and the fourth Euclidean distance, and weighting the similar blocks according to the weights corresponding to each similar block to obtain the noise reduction image corresponding to the binocular image.
In one embodiment, according to the first horizontal parallax matrix, a first euclidean distance of the first left-view guiding image in the first right-view guiding image and a second euclidean distance of the first right-view guiding image in the first left-view guiding image are calculated by using a non-local mean filtering algorithm; calculating a third Euclidean distance of the first left-view guiding image in the self image by using a non-local mean value filtering algorithm, and a fourth Euclidean distance of the first right-view guiding image in the self image comprises the following steps:
according to the first horizontal parallax matrix, non-local average filtering is carried out on the first left-view guiding image in a first image block taking a horizontal parallax pixel as a center in the first right-view guiding image, so that the first Euclidean distance is obtained;
according to the first horizontal parallax matrix, performing non-local mean filtering on the first right-view guiding image in a second image block taking a horizontal parallax pixel as a center in the first left-view guiding image to obtain the second Euclidean distance;
non-local mean filtering is carried out on a third image block taking a pixel to be processed as a center in the first left-view guiding image in the self image, so that the third Euclidean distance is obtained;
And carrying out non-local mean filtering on the first right-view guiding image in a fourth image block taking a pixel to be processed as a center in the self image to obtain the fourth Euclidean distance.
In one embodiment, the calculating, by using a non-local mean filtering algorithm, the first horizontal parallax matrix corresponding to the shallowest layer guiding image in the multi-layer guiding image includes:
calculating a second horizontal parallax matrix corresponding to the deepest guiding image in the multi-layer guiding images by using a non-local mean filtering algorithm;
determining an initial horizontal parallax matrix corresponding to a guide image of the upper layer of the deepest guide image according to the second horizontal parallax matrix;
optimizing an initial horizontal parallax matrix corresponding to a guide image of the upper layer of the deepest guide image by using a non-local mean value filtering algorithm;
repeating the determining step and the optimizing step until the optimization of the initial horizontal parallax matrix corresponding to the shallowest layer guiding image is completed, and obtaining the first horizontal parallax matrix.
In one embodiment, the deepest guide image includes a second left view guide image and a second right view guide image; the calculating of the second horizontal parallax matrix corresponding to the deepest guiding image in the multi-layer guiding image by using the non-local mean filtering algorithm comprises the following steps:
Determining a fifth image block by taking a preset first pixel point in the second left-view guiding image as a center;
determining a first search window of the second right view guide image;
searching a first similar block corresponding to the fifth image block in the first search window;
obtaining a second left-view parallax matrix according to the horizontal position difference of the central pixel point of the first similar block in the second left-view guiding image and the second right-view guiding image;
determining a sixth image block by taking a preset second pixel point in the second right-view guiding image as a center;
determining a second search window of the second left-view guide image;
searching a second similar block corresponding to the sixth image block in the second search window;
obtaining a second right-view parallax matrix according to the horizontal position difference of the central pixel point of the second similar block in the second left-view guiding image and the second right-view guiding image;
and obtaining the second horizontal parallax matrix according to the second left-view parallax matrix and the second right-view parallax matrix.
In one embodiment, obtaining the second left-view parallax matrix according to the horizontal position difference between the second left-view guiding image and the second right-view guiding image of the center pixel point of the first similar block includes:
Obtaining an initial left-view parallax matrix according to the horizontal position difference of the central pixel points of the first similar blocks in the second left-view guiding image and the second right-view guiding image;
adjusting the first search window according to the magnitude relation between the parallax difference value between the preset third pixel point, the preset fourth pixel point and the preset fifth pixel point in the initial left-view parallax matrix and the preset first threshold value;
searching a first similar block corresponding to the third image block in the adjusted first search window, and repeating the adjustment steps until the boundary of the second right-view guiding image is searched, and obtaining a second left-view parallax matrix according to the horizontal position difference of the central pixel points of all the first similar blocks in the second left-view guiding image and the second right-view guiding image.
In one embodiment, obtaining the second horizontal parallax matrix according to the second left-view parallax matrix and the second right-view parallax matrix includes:
judging whether the parallax difference value of each pixel point in the second left-view parallax matrix and the second right-view parallax matrix is smaller than a preset second threshold value or not;
When the parallax difference value of each pixel point in the second left-view parallax matrix and the second right-view parallax matrix is smaller than the second threshold value, obtaining a second horizontal parallax matrix;
and when error pixels with parallax difference values larger than or equal to the second threshold value exist in the second left-view parallax matrix and the second right-view parallax matrix, adjusting the numerical value corresponding to the error pixels in the second horizontal parallax matrix to be a preset parallax threshold value.
In one embodiment, performing a multi-layer downsampling process and a noise reduction process on the binocular image to obtain a multi-layer guiding image includes:
performing multi-layer downsampling processing on the binocular image to obtain a multi-layer downsampled image;
and respectively carrying out noise reduction treatment on each layer of downsampled image to obtain a guide image corresponding to each layer of downsampled image.
In a second aspect, an embodiment of the present application provides a noise reduction apparatus for binocular images, including:
the acquisition module is used for acquiring binocular images, wherein the binocular images comprise left-view images and right-view images; the preprocessing module is used for carrying out multi-layer downsampling processing and noise reduction processing on the binocular image to obtain a multi-layer guide image; the first calculation module is used for calculating a first horizontal parallax matrix corresponding to the shallowest layer guide image in the multilayer guide images by using a non-local mean value filtering algorithm; the shallowest layer guiding image comprises a first left-view guiding image and a first right-view guiding image; the second calculation module is used for calculating a first Euclidean distance of the first left-view guiding image in the first right-view guiding image and a second Euclidean distance of the first right-view guiding image in the first left-view guiding image by utilizing a non-local mean value filtering algorithm according to the first horizontal parallax matrix; calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by using a non-local mean filtering algorithm; and the noise reduction module is used for determining a plurality of similar blocks in the left-view image and the right-view image and weights corresponding to the similar blocks according to the first Euclidean distance, the second Euclidean distance, the third Euclidean distance and the fourth Euclidean distance, and carrying out weighting processing on the similar blocks according to the weights corresponding to the similar blocks to obtain a noise reduction image corresponding to the binocular image.
In a third aspect, an embodiment of the present application further provides an electronic device, including a memory, and a processor, where the memory stores a computer program, and the processor is configured to run the computer program to perform the method for denoising a binocular image according to the first aspect.
In a fourth aspect, an embodiment of the present application further provides a storage medium, where a computer program is stored, where the computer program, when executed by a processor, implements a method for denoising a binocular image according to the first aspect.
Compared with the related art, the method, the device, the electronic device and the storage medium for denoising the binocular image provided by the embodiment of the application further perform multi-layer downsampling processing and denoising processing on the binocular image by acquiring the binocular image to obtain a multi-layer guide image; calculating to obtain a first horizontal parallax matrix corresponding to the shallowest layer guiding image in the multi-layer guiding image by using a non-local mean filtering algorithm; the shallowest layer guiding image comprises a first left-view guiding image and a first right-view guiding image; according to the first horizontal parallax matrix, calculating a first Euclidean distance of the first left-view guiding image in the first right-view guiding image and a second Euclidean distance of the first right-view guiding image in the first left-view guiding image by utilizing a non-local mean filtering algorithm; calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by using a non-local mean filtering algorithm; and finally, determining a plurality of similar blocks in the left-view image and the right-view image and weights corresponding to the similar blocks according to the first Euclidean distance, the second Euclidean distance, the third Euclidean distance and the fourth Euclidean distance, and carrying out weighting processing on the plurality of similar blocks according to the weights corresponding to the similar blocks to obtain the noise reduction image of the corresponding binocular image. The application solves the problem of larger image noise acquired by the endoscope in the related technology, and realizes the technical effect of reducing the image noise of the binocular image acquired by the binocular endoscope.
The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below to provide a more thorough understanding of the other features, objects, and advantages of the application.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute a limitation on the application. In the drawings:
FIG. 1 is a flow chart of a method of denoising a binocular image according to an embodiment of the present application;
FIG. 2 is a flow chart of binocular endoscopic stereoscopic imaging;
fig. 3 is a block diagram of a structure of a noise reduction device of a binocular image according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The present application will be described and illustrated with reference to the accompanying drawings and examples in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application. All other embodiments, which can be made by a person of ordinary skill in the art based on the embodiments provided by the present application without making any inventive effort, are intended to fall within the scope of the present application. Moreover, it should be appreciated that while such a development effort might be complex and lengthy, it would nevertheless be a routine undertaking of design, fabrication, or manufacture for those of ordinary skill having the benefit of this disclosure, and thus should not be construed as having the benefit of this disclosure.
Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is to be expressly and implicitly understood by those of ordinary skill in the art that the described embodiments of the application can be combined with other embodiments without conflict.
Unless defined otherwise, technical or scientific terms used herein should be given the ordinary meaning as understood by one of ordinary skill in the art to which this application belongs. The terms "a," "an," "the," and similar referents in the context of the application are not to be construed as limiting the quantity, but rather as singular or plural. The terms "comprising," "including," "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to only those steps or elements but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. The terms "connected," "coupled," and the like in connection with the present application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as used herein means greater than or equal to two. "and/or" describes an association relationship of an association object, meaning that there may be three relationships, e.g., "a and/or B" may mean: a exists alone, A and B exist together, and B exists alone. The terms "first," "second," "third," and the like, as used herein, are merely distinguishing between similar objects and not representing a particular ordering of objects.
The present embodiment provides a method for denoising a binocular image, and fig. 1 is a flowchart of a method for denoising a binocular image according to an embodiment of the present application, as shown in fig. 1, the method includes:
step S101, a binocular image is acquired, wherein the binocular image comprises a left view image and a right view image.
In this embodiment, the binocular image may be acquired by a binocular endoscope, and the left view image P1 and the right view image P2 are images after stereoscopic correction.
In the present embodiment, the parallax range of the binocular image [ -Th1, th1] may also be acquired.
And step S102, performing multi-layer downsampling and noise reduction on the binocular image to obtain a multi-layer guide image.
In this embodiment, in order to achieve both high noise reduction effect and low computational resource consumption, downsampling and noise reduction of 2 layers or less may be performed on the binocular image to obtain the guide image P1 i ，P2 i (i=0, …, N, n.ltoreq.2), the right and left images downsampled in the same layer can be subjected to the noise reduction processing to the same extent, respectively.
Step S103, calculating to obtain a first horizontal parallax matrix corresponding to the shallowest layer guiding image in the multi-layer guiding images by using a non-local mean filtering algorithm; the shallowest layer guide image includes a first left view guide image and a first right view guide image.
Referring to the above embodiment, the left-view image P1 is subjected to the multi-layer downsampling process and the noise reduction process, so that the shallowest layer guiding image of the left-view image P1, i.e., the first left-view guiding image P1, can be obtained 0 . Similarly, the right-view image P2 is subjected to multi-layer downsampling and noise reduction to the same extent, so as to obtain the shallowest layer guiding image of the right-view image P2, i.e. the first right-view guiding image P2 0 。
Step S104, according to the first horizontal parallax matrix, calculating a first Euclidean distance of the first left-view guiding image in the first right-view guiding image and a second Euclidean distance of the first right-view guiding image in the first left-view guiding image by utilizing a non-local mean filtering algorithm; and calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by using a non-local mean filtering algorithm.
In this embodiment, the shallowest layer guiding image includes a first left view guidingImage P1 0 And a first right-view guidance image P2 0 The method comprises the steps of carrying out a first treatment on the surface of the So according to the first horizontal parallax matrix, the first left-view guiding image P1 is calculated by utilizing a non-local mean filtering algorithm 0 At the first right-view guidance image P2 0 The first right-view guiding image P2 0 At the first left-view guiding image P1 0 A second Euclidean distance of (a); calculating to obtain the first left-view guiding image P1 by utilizing a non-local mean filtering algorithm 0 A third Euclidean distance in the self image, and the first right view guiding image P2 0 The fourth Euclidean distance in the self image is realized by the following steps:
step one, according to the first horizontal parallax matrix, the first left-view guiding image P1 is subjected to 0 At the first right-view guidance image P2 0 Non-local mean filtering is carried out on a first image block taking a horizontal parallax pixel as a center, and the first Euclidean distance is obtained;
step two, according to the first horizontal parallax matrix, the first right-view guiding image P2 is subjected to 0 At the first left-view guiding image P1 0 Non-local mean filtering is carried out on a second image block taking the horizontal parallax pixel as the center, and the second Euclidean distance is obtained;
step three, regarding the first left-view guiding image P1 0 Non-local mean filtering is carried out on a third image block taking a pixel to be processed as a center in the self image, so that the third Euclidean distance is obtained;
Fourth, the first right-view guiding image P2 is displayed 0 And carrying out non-local mean filtering in a fourth image block taking a pixel to be processed as a center in the self image to obtain the fourth Euclidean distance.
Step S105, determining a plurality of similar blocks in the left view image and the right view image and weights corresponding to each similar block according to the first euclidean distance, the second euclidean distance, the third euclidean distance and the fourth euclidean distance, and weighting the plurality of similar blocks according to the weights corresponding to each similar block to obtain a noise reduction image corresponding to the binocular image.
In the above embodiment, the weight calculation may be performed based on the first euclidean distance, the second euclidean distance, the third euclidean distance, and the fourth euclidean distance, or the first N similar blocks with smaller distances may be obtained according to the first euclidean distance, the second euclidean distance, the third euclidean distance, and the fourth euclidean distance, and then the weight calculation may be performed.
In this embodiment, the matching processing may be performed by using the left-view image and the right-view image that are captured simultaneously, and since the similarity between the left-view image and the right-view image is high, the noise reduction processing is performed on the binocular image by using the non-local mean filtering algorithm, so that the edge protection and the noise reduction effect may be improved.
Meanwhile, after the binocular image is subjected to downsampling, the parallax of the acquired guide image is reduced, the matching range of the similar blocks is reduced, and the computing resources can be saved.
Fig. 2 is a flowchart of stereoscopic imaging of a binocular endoscope, as shown in fig. 2, the binocular endoscope needs to perform correction and image amplification operations before performing 3D image stereoscopic display, and the noise reduction method of the binocular image provided in this embodiment may be performed after correction, so that the noise reduction processing of the binocular image may be performed by using the characteristics that similar textures and parallax ranges exist in the left view image and the right view image, and the noise reduction processing of the binocular image may be performed before image amplification, so that image noise may be reduced and partial image noise may be prevented from being amplified.
Through the steps S101 to S105, the multi-layer downsampling process and the noise reduction process are further performed on the binocular image by obtaining the binocular image, so as to obtain a multi-layer guiding image; calculating to obtain a first horizontal parallax matrix corresponding to the shallowest layer guiding image in the multi-layer guiding image by using a non-local mean filtering algorithm; the shallowest layer guiding image comprises a first left-view guiding image and a first right-view guiding image; according to the first horizontal parallax matrix, calculating to obtain a first Euclidean distance of a first left-view guiding image in a first right-view guiding image by using a non-local mean value filtering algorithm, calculating to obtain a third Euclidean distance of the first left-view guiding image in a self image by using a non-local mean value filtering algorithm, and calculating to obtain a fourth Euclidean distance of the first right-view guiding image in the self image; and finally, determining a plurality of similar blocks in the left-view image and the right-view image and weights corresponding to the similar blocks according to the first Euclidean distance, the second Euclidean distance, the third Euclidean distance and the fourth Euclidean distance, and carrying out weighting processing on the plurality of similar blocks according to the weights corresponding to the similar blocks to obtain the noise reduction image of the corresponding binocular image. The application solves the problem of larger image noise acquired by the endoscope in the related technology, and realizes the technical effect of reducing the image noise of the binocular image acquired by the binocular endoscope.
In some embodiments, the multi-layer downsampling and denoising processes are performed on the binocular image, so that a multi-layer guide image is obtained through the following steps:
and step 1, carrying out multi-layer downsampling processing on the binocular image to obtain a multi-layer downsampled image.
And 2, respectively carrying out noise reduction treatment on each layer of downsampled image to obtain a guide image corresponding to each layer of downsampled image.
In this embodiment, in order to achieve both high noise reduction effect and low computational resource consumption, downsampling processing of 2 layers or less may be performed on the binocular image, and noise reduction processing of the same level may be performed on the left and right images downsampled on the same layer, respectively, to obtain the guide image P1 i ，P2 i (i=0, …, N, n.ltoreq.2), where P1 i Is a left-view guiding image, P2 i The method is a right-view guide image, after downsampling is carried out on the binocular image, the parallax of the acquired guide image is reduced, the matching range of similar blocks is reduced, and the calculation resources can be saved.
In some of these embodiments, the deepest guide image includes a second left-view guide image P1 n And a second right-view guide image P2 n The method comprises the steps of carrying out a first treatment on the surface of the Calculating to obtain the shallowest layer guiding image corresponding to the multi-layer guiding image by using a non-local mean filtering algorithm Is realized by the following steps:
and step 1, calculating to obtain a second horizontal parallax matrix corresponding to the deepest layer guiding image in the multi-layer guiding images by using a non-local mean filtering algorithm.
In this embodiment, the calculation of the second horizontal parallax matrix corresponding to the deepest guiding image in the multi-layer guiding image by using the non-local mean filtering algorithm is implemented by the following steps:
step 1 1 And determining a fifth image block by taking a preset first pixel point in the second left-view guiding image as a center.
Step 1 2 And determining a first search window of the second right-view guide image.
In this embodiment, the search window may be adaptively adjusted according to the matched second left-view parallax matrix diff_l and the edge image of the deepest guide image.
Step 1 3 And searching a first similar block corresponding to the fifth image block in the first search window.
Step 1 4 And obtaining a second left-view parallax matrix according to the horizontal position difference of the central pixel point of the first similar block in the second left-view guiding image and the second right-view guiding image.
In this embodiment, the first pixel point P1 may be used n (x, y) centering on a matching block of size a x a and on the second right view guide image P2 n The first similar blocks for carrying out the self-adaptive adjustment of the search window are matched, and the central pixel point of each first similar block is positioned in the second left-view guiding image P1 n And the second right-view guidance image P2 n The horizontal position differences of (2) constitute a second left-view disparity matrix diff_l.
In the present embodiment, the parallax range of the second left-view parallax matrix diff_L is [ -Thr1/2 n ，Thr1/2 n ]And 2 x Thr1/2 n Wherein, [ -Thr1/2 n ，Thr1/2 n ]Is the allowable parallax range, 2 x Thr1/2 n Is the disparity that did not search for a qualified similar block.
Step 1 5 And determining a sixth image block by taking a preset second pixel point in the second right-view guiding image as a center.
Step 1 6 And determining a second search window of the second left-view guide image.
Step 1 7 And searching a second similar block corresponding to the sixth image block in the second search window.
Step 1 8 And obtaining a second right-view parallax matrix according to the horizontal position difference of the center pixel point of the second similar block in the second left-view guiding image and the second right-view guiding image.
In the present embodiment, the second right-view parallax matrix diff_r can be acquired similarly by the method of acquiring the second left-view parallax matrix diff_l described above.
Step 1 9 And obtaining the second horizontal parallax matrix according to the second left-view parallax matrix and the second right-view parallax matrix.
In this embodiment, obtaining the second horizontal parallax matrix according to the second left-view parallax matrix and the second right-view parallax matrix includes: judging whether the parallax difference value of each pixel point in the second left-view parallax matrix and the second right-view parallax matrix is smaller than a preset second threshold value or not; when the parallax difference value of each pixel point in the second left-view parallax matrix and the second right-view parallax matrix is smaller than the second threshold value, obtaining a second horizontal parallax matrix; and when error pixels with parallax difference values larger than or equal to the second threshold value exist in the second left-view parallax matrix and the second right-view parallax matrix, adjusting the numerical value corresponding to the error pixels in the second horizontal parallax matrix to be a preset parallax threshold value.
In this embodiment, after the second left-view parallax matrix diff_l and the second right-view parallax matrix diff_r are obtained, it may be further determined whether the parallax difference of each pixel point in the second left-view parallax matrix diff_l and the second right-view parallax matrix diff_r is smaller than a preset second threshold value Thr5, if the parallax difference of each pixel point is smaller than the second threshold value Thr5 If the similar block is not correctly matched, the second horizontal parallax matrix Diffn is obtained based on the second left-view parallax matrix diff_l and the second right-view parallax matrix diff_r, otherwise, the similar block is incorrectly matched, and the value corresponding to the wrong pixel point in the second horizontal parallax matrix Diffn can be adjusted to be a preset parallax threshold 2 x thr1/2 n 。
And 2, determining an initial horizontal parallax matrix corresponding to a guide image of the upper layer of the deepest guide image according to the second horizontal parallax matrix.
In the present embodiment, it is possible to guide the image P according to the deepest layer n Determining a top layer guide image P of the deepest layer guide image based on a second horizontal parallax matrix Diffn of the image n-1 Corresponding initial horizontal parallax matrix Diff n-1 Diffn=2, and determines the search window as: search method range (x，y)＝a*[Diff n-1 (x，y)-Thr6，Diff n-1 (x，y)+Thr6]Wherein Thr6 is the maximum threshold allowed by the preset parallax change.
And 3, optimizing an initial horizontal parallax matrix corresponding to a guide image of the upper layer of the deepest guide image by using a non-local mean filtering algorithm.
In this embodiment, the non-local mean filtering algorithm may be used to calculate the euclidean distance in the search window, and refer to the neighboring disparities for the initial horizontal disparity matrix Diff n-1 And (5) optimizing.
And 4, repeating the determining step and the optimizing step until the optimization of the initial horizontal parallax matrix corresponding to the shallowest layer guiding image is completed, and obtaining the first horizontal parallax matrix.
In this embodiment, the initial horizontal parallax matrices corresponding to the 0-n-1 layer guiding images can be optimized by referring to the above steps 2 and 3 until the shallowest layer guiding image P is completed 0 Optimizing the corresponding initial horizontal parallax matrix to obtain the first horizontal parallax matrix Diff 0 。
In some embodiments, according to the horizontal position difference between the second left-view guiding image and the second right-view guiding image of the central pixel point of the first similar block, the obtaining of the second left-view parallax matrix is implemented by the following steps:
step 1, obtaining an initial left-view parallax matrix according to the horizontal position difference of the central pixel points of the first similar blocks in the second left-view guiding image and the second right-view guiding image.
And 2, adjusting the first search window according to the magnitude relation between the parallax difference value between the preset third pixel point, the preset fourth pixel point and the preset fifth pixel point in the initial left-view parallax matrix and the preset first threshold value.
And 3, searching a first similar block corresponding to the third image block in the adjusted first search window, and repeating the adjustment step until the boundary of the second right-view guiding image is searched, and obtaining a second left-view parallax matrix according to the horizontal position difference of the central pixel points of all the first similar blocks in the second left-view guiding image and the second right-view guiding image.
In this embodiment, after matching of a part of the first similar blocks is completed, an initial left-view parallax matrix may be obtained, and the first search window may be adaptively adjusted according to the matched parallaxes in the initial left-view parallax matrix, so as to reduce the influence of image noise.
In the present embodiment, the second left-view guidance image P1 may also be respectively n And the second right-view guidance image P2 n And performing Edge detection to obtain Edge images Edge1 and Edge2.
Can be based on the magnitude relation between the parallax difference between the preset third pixel point diff_l (x-1, y) in the initial left-view parallax matrix diff_l, the preset fourth pixel point diff_l (x-1, y+1) and the preset fifth pixel point diff_l (x, y-1) and the preset first threshold value Thr2, and whether the first pixel point P1 exists in the Edge image Edge1 n Pixel point Edge1 (x, y) corresponding to (x, y) for searching the first search window range And (5) adjusting.
In the present embodiment, if the parallax difference between diff_l (x-1, y), diff_l (x-1, y+1), and diff_l (x, y-1) is smaller than Thr2, and diff_l (x-1, y), diff_l (x-1, y+1), and diff_l (x, y-1) are positive numbers or negative numbers, edge1 (x, y) =0, the maximum search range max_range and the minimum search range min_range can be determined:
max-range＝max(Diff_L(x-1，y)，Diff_L(x-1，y+1)，Diff_L(x，y-1))；
min-range＝min(Diff_L(x-1，y)，Diff_L(x-1，y+1)，Diff_L(x，y-1))；
if Diff_L (x-1, y), diff_L (x-1, y+1), and Diff_L (x, y-1) are positive numbers, then the first search window search range ＝a*[x+min-range，x+max-range+Thr3]The method comprises the steps of carrying out a first treatment on the surface of the If Diff_L (x-1, y), diff_L (x-1, y+1), and Diff_L (x, y-1) are negative numbers, then the first search window search range ＝a*[x+min-range-Thr3，x+max-range]。
If the first search window search is range The minimum Euclidean distance and the slightly larger Euclidean distance in the image processing device are smaller than Thr4, and the parallax of the slightly larger Euclidean distance is more similar to Diff_L (x-1, y), diff_L (x-1, y+1) and Diff_L (x, y-1), if the pixel point with the slightly larger Euclidean distance is taken as the parallax point, otherwise, the pixel point with the minimum Euclidean distance is taken as the parallax; if not, the parallax is 2 x Thr1/2n, wherein Thr3 is a preset maximum threshold value of the parallax change of the adjacent pixel points when no obvious depth change exists, thr4 is a preset second threshold value, and the adjacent pixel points are considered as similar blocks when the Euclidean distance is smaller than Thr 4.
In the present embodiment, if the parallax difference between diff_l (x-1, y), diff_l (x-1, y+1), and diff_l (x, y-1) is smaller than Thr2, and the values of diff_l (x-1, y), diff_l (x-1, y+1), and diff_l (x, y-1) are positive and negative, edge1 (x, y) =0, the maximum search range max_range and the minimum search range min_range can be determined:
max-range＝max(Diff_L(x-1，y)，Diff_L(x-1，y+1)，Diff_L(x，y-1))；
min-range＝min(Diff_L(x-1，y)，Diff_L(x-1，y+1)，Diff_L(x，y-1))；
first search window search range ＝a*[x+min-range-Thr3，x+max-range+Thr3]。
If the first search window search is range The minimum Euclidean distance and the slightly larger Euclidean distance in the image processing device are smaller than Thr4, and the parallax of the slightly larger Euclidean distance is more similar to Diff_L (x-1, y), diff_L (x-1, y+1) and Diff_L (x, y-1), if the pixel point with the slightly larger Euclidean distance is taken as the parallax point, otherwise, the pixel point with the minimum Euclidean distance is taken as the parallax; if not, the parallax is noted as 2 x Thr1/2n.
In the present embodiment, if the parallax difference between diff_l (x-1, y), diff_l (x-1, y+1) and diff_l (x, y-1) is smaller than Thr2, and Edge1 (x, y) =1, the corresponding first search window search may be determined according to the positive and negative of diff_l (x-1, y), diff_l (x-1, y+1) and diff_l (x, y-1) using the above method range And search in the first search window range Matching of similar blocks is performed.
If the first search window search is range The minimum Euclidean distance and the slightly larger Euclidean distance in the image processing device are smaller than Thr4, and the parallax of the slightly larger Euclidean distance is more similar to Diff_L (x-1, y), diff_L (x-1, y+1) and Diff_L (x, y-1), if the pixel point with the slightly larger Euclidean distance is taken as the parallax point, otherwise, the pixel point with the minimum Euclidean distance is taken as the parallax; if not, enlarging the first search window search range ＝a*[x-Thr1/2 n ，x+Thr1/2 n ]If the expanded first search window search range If the minimum Euclidean distance in the pixel is smaller than Thr4, the pixel where the minimum Euclidean distance is located is taken as a parallax point, and if the minimum Euclidean distance is not present, the parallax is recorded as 2 x Thr1/2n.
In the present embodiment, if the parallax difference between diff_l (x-1, y), diff_l (x-1, y+1), and diff_l (x, y-1) is greater than Thr2, the first search window search is determined range ＝a*[x-Thr1/2 n ，x+Thr1/2 n ]。
If the first search window search is range If the minimum Euclidean distance in the pixel is smaller than Thr4, the pixel where the minimum Euclidean distance is located is taken as a parallax point, and if the minimum Euclidean distance is not present, the parallax is recorded as 2 x Thr1/2n.
In this embodiment, when matching the similar block and the parallax point, the direction and the size of the search window may be defined according to the matched parallax point, and the size of the search window may be adaptively adjusted, so as to further reduce the influence of image noise.
It should be noted that, in the embodiment provided by the present application, the similarity between the matching blocks is calculated by specifically using the block matching algorithm part in the non-local mean filtering algorithm. The degree of similarity between the calculated matching blocks is not limited to the method listed in the above embodiment, and may be implemented by other existing algorithms.
The present embodiment provides a noise reduction device for binocular images, and fig. 3 is a block diagram of the structure of the noise reduction device for binocular images according to an embodiment of the present application, as shown in fig. 3, the device includes: an obtaining module 302, configured to obtain a binocular image, where the binocular image includes a left view image and a right view image; the preprocessing module 304 is configured to perform multi-layer downsampling processing and noise reduction processing on the binocular image to obtain a multi-layer guide image; a first calculation module 306, configured to calculate, using a non-local mean filtering algorithm, a first horizontal parallax matrix corresponding to a shallowest layer guiding image in the multi-layer guiding image; the shallowest layer guiding image comprises a first left-view guiding image and a first right-view guiding image; a second calculation module 308, configured to calculate, according to the first horizontal parallax matrix, a first euclidean distance of the first left-view guiding image in the first right-view guiding image and a second euclidean distance of the first right-view guiding image in the first left-view guiding image by using a non-local mean filtering algorithm; the method is also used for calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by utilizing a non-local mean filtering algorithm; the noise reduction module 310 is configured to determine a plurality of similar blocks in the left view image and the right view image and weights corresponding to each similar block according to the first euclidean distance, the second euclidean distance, the third euclidean distance, and the fourth euclidean distance, and perform weighting processing on the plurality of similar blocks according to the weights corresponding to each similar block, so as to obtain a noise reduction image corresponding to the binocular image.
In some embodiments, the second calculating module 308 is further configured to perform non-local mean filtering on the first left-view guiding image in a first image block centered on a horizontal parallax pixel in the first right-view guiding image according to the first horizontal parallax matrix, to obtain the first euclidean distance; according to the first horizontal parallax matrix, performing non-local mean filtering on the first right-view guiding image in a second image block taking a horizontal parallax pixel as a center in the first left-view guiding image to obtain the second Euclidean distance; non-local mean filtering is carried out on a third image block taking a pixel to be processed as a center in the first left-view guiding image in the self image, so that the third Euclidean distance is obtained; and carrying out non-local mean filtering on the first right-view guiding image in a fourth image block taking a pixel to be processed as a center in the self image to obtain the fourth Euclidean distance.
In some embodiments, the first calculating module 306 is further configured to calculate, using a non-local mean filtering algorithm, a second horizontal parallax matrix corresponding to a deepest guiding image in the multi-layer guiding images; determining an initial horizontal parallax matrix corresponding to a guide image of the upper layer of the deepest guide image according to the second horizontal parallax matrix; optimizing an initial horizontal parallax matrix corresponding to a guide image of the upper layer of the deepest guide image by using a non-local mean value filtering algorithm; repeating the determining step and the optimizing step until the optimization of the initial horizontal parallax matrix corresponding to the shallowest layer guiding image is completed, and obtaining the first horizontal parallax matrix.
In some of these embodiments, the deepest guide image includes a second left view guide image and a second right view guide image; the first computing module 306 is further configured to determine a fifth image block centered on a preset first pixel point in the second left-view guiding image; determining a first search window of the second right view guide image; searching a first similar block corresponding to the fifth image block in the first search window; obtaining a second left-view parallax matrix according to the horizontal position difference of the central pixel point of the first similar block in the second left-view guiding image and the second right-view guiding image; determining a sixth image block by taking a preset second pixel point in the second right-view guiding image as a center; determining a second search window of the second left-view guide image; searching a second similar block corresponding to the sixth image block in the second search window; obtaining a second right-view parallax matrix according to the horizontal position difference of the central pixel point of the second similar block in the second left-view guiding image and the second right-view guiding image; and obtaining the second horizontal parallax matrix according to the second left-view parallax matrix and the second right-view parallax matrix.
In some of these embodiments, the first computing module 306 is further configured to obtain an initial left-view disparity matrix from a horizontal position gap of the second left-view guide image and the second right-view guide image at a center pixel point of the plurality of first similar blocks; adjusting the first search window according to the magnitude relation between the parallax difference value between the preset third pixel point, the preset fourth pixel point and the preset fifth pixel point in the initial left-view parallax matrix and the preset first threshold value; searching a first similar block corresponding to the third image block in the adjusted first search window, and repeating the adjustment steps until the boundary of the second right-view guiding image is searched, and obtaining a second left-view parallax matrix according to the horizontal position difference of the central pixel points of all the first similar blocks in the second left-view guiding image and the second right-view guiding image.
In some embodiments, the first computing module 306 is further configured to determine whether the parallax difference of each pixel in the second left-view parallax matrix and the second right-view parallax matrix is less than a preset second threshold; when the parallax difference value of each pixel point in the second left-view parallax matrix and the second right-view parallax matrix is smaller than the second threshold value, obtaining a second horizontal parallax matrix; and when error pixels with parallax difference values larger than or equal to the second threshold value exist in the second left-view parallax matrix and the second right-view parallax matrix, adjusting the numerical value corresponding to the error pixels in the second horizontal parallax matrix to be a preset parallax threshold value.
In some of these embodiments, the preprocessing module 304 is further configured to perform a multi-layer downsampling process on the binocular image to obtain a multi-layer downsampled image; and respectively carrying out noise reduction treatment on each layer of downsampled image to obtain a guide image corresponding to each layer of downsampled image.
It should be noted that, specific examples in this embodiment may refer to examples described in the foregoing embodiments and alternative implementations, and this embodiment is not repeated herein.
The present embodiment also provides an electronic device, fig. 4 is a schematic diagram of a hardware structure of the electronic device according to an embodiment of the present application, and as shown in fig. 4, the electronic device includes a memory 404 and a processor 402, where the memory 404 stores a computer program, and the processor 402 is configured to run the computer program to perform steps in any one of the method embodiments described above.
In particular, the processor 402 may include a Central Processing Unit (CPU), or an application specific integrated circuit (Application Specific Integrated Circuit, abbreviated as ASIC), or may be configured as one or more integrated circuits that implement embodiments of the present application.
Memory 404 may include, among other things, mass storage for data or instructions. By way of example, and not limitation, memory 404 may comprise a Hard Disk Drive (HDD), floppy Disk Drive, solid state Drive (Solid State Drive, SSD), flash memory, optical Disk, magneto-optical Disk, tape, or universal serial bus (Universal Serial Bus, USB) Drive, or a combination of two or more of the foregoing. Memory 404 may include removable or non-removable (or fixed) media, where appropriate. The memory 404 may be internal or external to the faulty image generation device, where appropriate. In a particular embodiment, the memory 404 is a Non-Volatile (Non-Volatile) memory. In a particular embodiment, the Memory 404 includes Read-Only Memory (ROM) and random access Memory (Random Access Memory, RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (Programmable Read-Only Memory, abbreviated PROM), an erasable PROM (Erasable Programmable Read-Only Memory, abbreviated EPROM), an electrically erasable PROM (Electrically Erasable Programmable Read-Only Memory, abbreviated EEPROM), an electrically rewritable ROM (Electrically Alterable Read-Only Memory, abbreviated EAROM), or a FLASH Memory (FLASH), or a combination of two or more of these. The RAM may be Static Random-Access Memory (SRAM) or dynamic Random-Access Memory (Dynamic Random Access Memory DRAM), where the DRAM may be a fast page mode dynamic Random-Access Memory (Fast Page Mode Dynamic Random Access Memory FPMDRAM), extended data output dynamic Random-Access Memory (Extended Date Out Dynamic Random Access Memory EDODRAM), synchronous dynamic Random-Access Memory (Synchronous Dynamic Random-Access Memory SDRAM), or the like, as appropriate.
Memory 404 may be used to store or cache various data files that need to be processed and/or used for communication, as well as possible computer program instructions for execution by processor 402.
The processor 402 implements the noise reduction method of any of the binocular images of the above embodiments by reading and executing computer program instructions stored in the memory 404.
Optionally, the electronic apparatus may further include a transmission device 406 and an input/output device 408, where the transmission device 406 is connected to the processor 402 and the input/output device 408 is connected to the processor 402.
Alternatively, in the present embodiment, the above-mentioned processor 402 may be configured to execute the following steps by a computer program:
s1, acquiring binocular images, wherein the binocular images comprise left-view images and right-view images.
S2, performing multi-layer downsampling and noise reduction on the binocular image to obtain a multi-layer guide image.
S3, calculating to obtain a first horizontal parallax matrix corresponding to the shallowest layer guide image in the multilayer guide images by using a non-local mean filtering algorithm; the shallowest layer guide image includes a first left view guide image and a first right view guide image.
S4, according to the first horizontal parallax matrix, calculating a first Euclidean distance of the first left-view guiding image in the first right-view guiding image and a second Euclidean distance of the first right-view guiding image in the first left-view guiding image by using a non-local mean filtering algorithm; and calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by using a non-local mean filtering algorithm.
S5, determining a plurality of similar blocks in the left-view image and the right-view image and weights corresponding to the similar blocks according to the first Euclidean distance, the second Euclidean distance, the third Euclidean distance and the fourth Euclidean distance, and weighting the similar blocks according to the weights corresponding to the similar blocks to obtain the noise reduction image corresponding to the binocular image.
It should be noted that, specific examples in this embodiment may refer to examples described in the foregoing embodiments and alternative implementations, and this embodiment is not repeated herein.
In addition, in combination with the noise reduction method of the binocular image in the above embodiment, the embodiment of the present application may be implemented by providing a storage medium. The storage medium has a computer program stored thereon; the computer program, when executed by a processor, implements the method of denoising a binocular image of any of the above embodiments.
It should be understood by those skilled in the art that the technical features of the above embodiments may be combined in any manner, and for brevity, all of the possible combinations of the technical features of the above embodiments are not described, however, they should be considered as being within the scope of the description provided herein, as long as there is no contradiction between the combinations of the technical features.
The foregoing examples illustrate only a few embodiments of the application, which are described in greater detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.
Claims (9)
1. A method of denoising a binocular image, the method comprising:
acquiring binocular images, wherein the binocular images comprise left-view images and right-view images;
performing multi-layer downsampling and noise reduction on the binocular image to obtain a multi-layer guide image;
calculating to obtain a first horizontal parallax matrix corresponding to the shallowest layer guiding image in the multi-layer guiding images by using a non-local mean filtering algorithm; the shallowest layer guiding image comprises a first left-view guiding image and a first right-view guiding image;
According to the first horizontal parallax matrix, a first Euclidean distance of the first left-view guiding image in the first right-view guiding image and a second Euclidean distance of the first right-view guiding image in the first left-view guiding image are calculated by utilizing a non-local mean filtering algorithm; calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by using a non-local mean filtering algorithm;
determining a plurality of similar blocks in the left-view image and the right-view image and weights corresponding to the similar blocks according to the first Euclidean distance, the second Euclidean distance, the third Euclidean distance and the fourth Euclidean distance, and weighting the similar blocks according to the weights corresponding to the similar blocks to obtain a noise reduction image corresponding to the binocular image;
the calculating, by using a non-local mean filtering algorithm, a first horizontal parallax matrix corresponding to a shallowest layer guiding image in the multi-layer guiding image includes:
calculating a second horizontal parallax matrix corresponding to the deepest guiding image in the multi-layer guiding images by using a non-local mean filtering algorithm;
Determining an initial horizontal parallax matrix corresponding to a guide image of the upper layer of the deepest guide image according to the second horizontal parallax matrix;
calculating the Euclidean distance in a search window by using a non-local mean value filtering algorithm, and referring to adjacent parallaxes, optimizing the initial horizontal parallax matrix;
repeating the determining step and the optimizing step until the optimization of the initial horizontal parallax matrix corresponding to the shallowest layer guiding image is completed, and obtaining the first horizontal parallax matrix.
2. The method of claim 1, wherein calculating a first euclidean distance of the first left-view guide image in the first right-view guide image, a second euclidean distance of the first right-view guide image in the first left-view guide image, a third euclidean distance of the first left-view guide image in the self image, and a fourth euclidean distance of the first right-view guide image in the self image using a non-local mean filtering algorithm according to the first horizontal disparity matrix comprises:
according to the first horizontal parallax matrix, non-local average filtering is carried out on the first left-view guiding image in a first image block taking a horizontal parallax pixel as a center in the first right-view guiding image, so that the first Euclidean distance is obtained;
According to the first horizontal parallax matrix, performing non-local mean filtering on the first right-view guiding image in a second image block taking a horizontal parallax pixel as a center in the first left-view guiding image to obtain the second Euclidean distance;
non-local mean filtering is carried out on a third image block taking a pixel to be processed as a center in the first left-view guiding image in the self image, so that the third Euclidean distance is obtained;
and carrying out non-local mean filtering on the first right-view guiding image in a fourth image block taking a pixel to be processed as a center in the self image to obtain the fourth Euclidean distance.
3. The method of noise reduction of binocular images according to claim 2, wherein the deepest layer guide image includes a second left view guide image and a second right view guide image; the calculating of the second horizontal parallax matrix corresponding to the deepest guiding image in the multi-layer guiding image by using the non-local mean filtering algorithm comprises the following steps:
determining a fifth image block by taking a preset first pixel point in the second left-view guiding image as a center;
determining a first search window of the second right view guide image;
Searching a first similar block corresponding to the fifth image block in the first search window;
obtaining a second left-view parallax matrix according to the horizontal position difference of the central pixel point of the first similar block in the second left-view guiding image and the second right-view guiding image;
determining a sixth image block by taking a preset second pixel point in the second right-view guiding image as a center;
determining a second search window of the second left-view guide image;
searching a second similar block corresponding to the sixth image block in the second search window;
obtaining a second right-view parallax matrix according to the horizontal position difference of the central pixel point of the second similar block in the second left-view guiding image and the second right-view guiding image;
and obtaining the second horizontal parallax matrix according to the second left-view parallax matrix and the second right-view parallax matrix.
4. A method of denoising a binocular image according to claim 3, wherein obtaining a second left-view parallax matrix from a horizontal position difference between the second left-view guide image and the second right-view guide image of the center pixel point of the first similar block comprises:
Obtaining an initial left-view parallax matrix according to the horizontal position difference of the central pixel points of the first similar blocks in the second left-view guiding image and the second right-view guiding image;
adjusting the first search window according to the magnitude relation between the parallax difference value between the preset third pixel point, the preset fourth pixel point and the preset fifth pixel point in the initial left-view parallax matrix and the preset first threshold value;
searching a first similar block corresponding to the third image block in the adjusted first search window, and repeating the adjustment steps until the boundary of the second right-view guiding image is searched, and obtaining a second left-view parallax matrix according to the horizontal position difference of the central pixel points of all the first similar blocks in the second left-view guiding image and the second right-view guiding image.
5. A method of denoising a binocular image according to claim 3, wherein deriving the second horizontal disparity matrix from the second left-view disparity matrix and the second right-view disparity matrix comprises:
judging whether the parallax difference value of each pixel point in the second left-view parallax matrix and the second right-view parallax matrix is smaller than a preset second threshold value or not;
When the parallax difference value of each pixel point in the second left-view parallax matrix and the second right-view parallax matrix is smaller than the second threshold value, obtaining a second horizontal parallax matrix;
and when error pixels with parallax difference values larger than or equal to the second threshold value exist in the second left-view parallax matrix and the second right-view parallax matrix, adjusting the numerical value corresponding to the error pixels in the second horizontal parallax matrix to be a preset parallax threshold value.
6. The method of denoising a binocular image according to any one of claims 1 or 5, wherein performing a multi-layer downsampling process and a denoising process on the binocular image to obtain a multi-layer guide image comprises:
performing multi-layer downsampling processing on the binocular image to obtain a multi-layer downsampled image;
and respectively carrying out noise reduction treatment on each layer of downsampled image to obtain a guide image corresponding to each layer of downsampled image.
7. A noise reduction device for binocular images, the device comprising:
the acquisition module is used for acquiring binocular images, wherein the binocular images comprise left-view images and right-view images;
the preprocessing module is used for carrying out multi-layer downsampling processing and noise reduction processing on the binocular image to obtain a multi-layer guide image;
The first calculation module is used for calculating a first horizontal parallax matrix corresponding to the shallowest layer guide image in the multilayer guide images by using a non-local mean value filtering algorithm; the shallowest layer guiding image comprises a first left-view guiding image and a first right-view guiding image;
the second calculation module is used for calculating a first Euclidean distance of the first left-view guiding image in the first right-view guiding image and a second Euclidean distance of the first right-view guiding image in the first left-view guiding image by utilizing a non-local mean value filtering algorithm according to the first horizontal parallax matrix; calculating a third Euclidean distance of the first left-view guiding image in the self image and a fourth Euclidean distance of the first right-view guiding image in the self image by using a non-local mean filtering algorithm;
the noise reduction module is used for determining a plurality of similar blocks in the left-view image and the right-view image and weights corresponding to the similar blocks according to the first Euclidean distance, the second Euclidean distance, the third Euclidean distance and the fourth Euclidean distance, and carrying out weighting processing on the similar blocks according to the weights corresponding to the similar blocks to obtain a noise reduction image corresponding to the binocular image;
The first calculation module is further used for calculating a second horizontal parallax matrix corresponding to the deepest guiding image in the multi-layer guiding images by using a non-local mean filtering algorithm; determining an initial horizontal parallax matrix corresponding to a guide image of the upper layer of the deepest guide image according to the second horizontal parallax matrix; and calculating the Euclidean distance in a search window by using a non-local mean value filtering algorithm, and referring to the adjacent parallax, optimizing an initial horizontal parallax matrix corresponding to the guide image of the upper layer of the deepest guide image.
8. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, the processor being arranged to run the computer program to perform the method of denoising a binocular image according to any one of claims 1 to 6.
9. A storage medium having a computer program stored therein, wherein the computer program when executed by a processor implements the method of denoising a binocular image of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211378346.6A CN115908170B (en) | 2022-11-04 | 2022-11-04 | Noise reduction method and device for binocular image, electronic device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211378346.6A CN115908170B (en) | 2022-11-04 | 2022-11-04 | Noise reduction method and device for binocular image, electronic device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115908170A CN115908170A (en) | 2023-04-04 |
| CN115908170B true CN115908170B (en) | 2023-11-21 |
Family
ID=86492681
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211378346.6A Active CN115908170B (en) | 2022-11-04 | 2022-11-04 | Noise reduction method and device for binocular image, electronic device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115908170B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102045571A (en) * | 2011-01-13 | 2011-05-04 | 北京工业大学 | Fast iterative search algorithm for stereo video coding |
| CN102254348A (en) * | 2011-07-25 | 2011-11-23 | 北京航空航天大学 | Block matching parallax estimation-based middle view synthesizing method |
| CN103606132A (en) * | 2013-10-31 | 2014-02-26 | 西安电子科技大学 | Multiframe digital image denoising method based on space domain and time domain combination filtering |
| CN107025660A (en) * | 2016-02-01 | 2017-08-08 | 北京三星通信技术研究有限公司 | A kind of method and apparatus for determining binocular dynamic visual sensor image parallactic |
| CN107786780A (en) * | 2017-11-03 | 2018-03-09 | 深圳Tcl新技术有限公司 | Video image noise reducing method, device and computer-readable recording medium |
| CN108986197A (en) * | 2017-11-30 | 2018-12-11 | 成都通甲优博科技有限责任公司 | 3D skeleton line construction method and device |
| CN109146803A (en) * | 2018-07-26 | 2019-01-04 | 北京航空航天大学 | SAR image radiometric resolution method for improving and device based on multi-angle image |
| CN109448036A (en) * | 2019-01-23 | 2019-03-08 | 深兰人工智能芯片研究院（江苏）有限公司 | A kind of method and device determining disparity map based on binocular image |
| CN111553878A (en) * | 2020-03-23 | 2020-08-18 | 四川公路工程咨询监理有限公司 | Method for detecting paving uniformity of asphalt pavement mixture based on binocular vision |
| CN113034666A (en) * | 2021-02-01 | 2021-06-25 | 中国计量大学 | Stereo matching method based on pyramid parallax optimization cost calculation |
| WO2022052582A1 (en) * | 2020-09-08 | 2022-03-17 | 苏州浪潮智能科技有限公司 | Image registration method and device, electronic apparatus, and storage medium |
| CN114255268A (en) * | 2020-09-24 | 2022-03-29 | 武汉Tcl集团工业研究院有限公司 | Disparity map processing and deep learning model training method and related equipment |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170109867A1 (en) * | 2015-10-16 | 2017-04-20 | Motorola Mobility Llc | Camera array for performing non-local means image processing over multiple sequential images |
| US10643308B2 (en) * | 2017-10-11 | 2020-05-05 | Gopro, Inc. | Double non-local means denoising |
-
2022
- 2022-11-04 CN CN202211378346.6A patent/CN115908170B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102045571A (en) * | 2011-01-13 | 2011-05-04 | 北京工业大学 | Fast iterative search algorithm for stereo video coding |
| CN102254348A (en) * | 2011-07-25 | 2011-11-23 | 北京航空航天大学 | Block matching parallax estimation-based middle view synthesizing method |
| CN103606132A (en) * | 2013-10-31 | 2014-02-26 | 西安电子科技大学 | Multiframe digital image denoising method based on space domain and time domain combination filtering |
| CN107025660A (en) * | 2016-02-01 | 2017-08-08 | 北京三星通信技术研究有限公司 | A kind of method and apparatus for determining binocular dynamic visual sensor image parallactic |
| CN107786780A (en) * | 2017-11-03 | 2018-03-09 | 深圳Tcl新技术有限公司 | Video image noise reducing method, device and computer-readable recording medium |
| CN108986197A (en) * | 2017-11-30 | 2018-12-11 | 成都通甲优博科技有限责任公司 | 3D skeleton line construction method and device |
| CN109146803A (en) * | 2018-07-26 | 2019-01-04 | 北京航空航天大学 | SAR image radiometric resolution method for improving and device based on multi-angle image |
| CN109448036A (en) * | 2019-01-23 | 2019-03-08 | 深兰人工智能芯片研究院（江苏）有限公司 | A kind of method and device determining disparity map based on binocular image |
| CN111553878A (en) * | 2020-03-23 | 2020-08-18 | 四川公路工程咨询监理有限公司 | Method for detecting paving uniformity of asphalt pavement mixture based on binocular vision |
| WO2022052582A1 (en) * | 2020-09-08 | 2022-03-17 | 苏州浪潮智能科技有限公司 | Image registration method and device, electronic apparatus, and storage medium |
| CN114255268A (en) * | 2020-09-24 | 2022-03-29 | 武汉Tcl集团工业研究院有限公司 | Disparity map processing and deep learning model training method and related equipment |
| CN113034666A (en) * | 2021-02-01 | 2021-06-25 | 中国计量大学 | Stereo matching method based on pyramid parallax optimization cost calculation |
Non-Patent Citations (3)
| Title |
|---|
| Non-local Means for Stereo Image Denoising Using Structural Similarity;Monagi H. Alkinani等;《International Conference Image Analysis and Recognition》;51-59 * |
| 双目立体视觉稠密视差算法研究;刘小莉;《中国优秀硕士学位论文全文数据库_信息科技辑》;I138-699 * |
| 基于双目视觉的裸眼3D图像生成算法及技术研究;郑晓洁;《中国优秀硕士学位论文全文数据库_信息科技辑》;I138-263 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115908170A (en) | 2023-04-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pham et al. | Domain transformation-based efficient cost aggregation for local stereo matching | |
| CN104574342B (en) | The noise recognizing method and Noise Identification device of parallax depth image | |
| CN111402170B (en) | Image enhancement method, device, terminal and computer readable storage medium | |
| US20110128282A1 (en) | Method for Generating the Depth of a Stereo Image | |
| US20170270644A1 (en) | Depth image Denoising Method and Denoising Apparatus | |
| CN109974623B (en) | Three-dimensional information acquisition method and device based on line laser and binocular vision | |
| US20110026834A1 (en) | Image processing apparatus, image capture apparatus, image processing method, and program | |
| EP3582182A1 (en) | A method, a device, and a system for estimating a sub-pixel position of an extreme point in an image | |
| TWI502544B (en) | Disparity estimation method of stereoscopic image | |
| WO2020215257A1 (en) | Image stereo matching method and assisted driving apparatus | |
| US20230154025A1 (en) | Light field encoded imaging method and apparatus for scattering scene | |
| US11256949B2 (en) | Guided sparse feature matching via coarsely defined dense matches | |
| CN111739071A (en) | Rapid iterative registration method, medium, terminal and device based on initial value | |
| CN115908170B (en) | Noise reduction method and device for binocular image, electronic device and storage medium | |
| CN113034666B (en) | Stereo matching method based on pyramid parallax optimization cost calculation | |
| CN114331919B (en) | Depth recovery method, electronic device, and storage medium | |
| CN115661258A (en) | Calibration method and device, distortion correction method and device, storage medium and terminal | |
| CN116188554A (en) | Three-dimensional imaging method and device based on binocular stereo measurement endoscope | |
| CN115176459A (en) | Virtual viewpoint synthesis method, electronic device, and computer-readable medium | |
| CN113538538A (en) | Binocular image alignment method, electronic device, and computer-readable storage medium | |
| CN112150532A (en) | Image processing method and device, electronic equipment and computer readable medium | |
| US10354400B2 (en) | Method and system for matching stereo images | |
| CN112765390A (en) | Stereo matching method with double search intervals | |
| JP2021068356A (en) | Imaging apparatus, method and program | |
| CN111383185B (en) | Hole filling method based on dense parallax map and vehicle-mounted equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |